Elements Of Data Visualization Final Project: Cancer Data Analysis
By Alex Cortez, Adam Hernandez, and Alex Pearce
GitHub: https://github.com/ACollectionOfAtoms/DV_FinalProject
This data was provided by the Enigma.io database, which can be found here. It is described as such: United States Cancer Statistics (USCS) information is aggregated from thousands of local cancer registries and published by the National Program of Cancer Registries (NPCR), a division of the Centers for Disease Control and Prevention (CDC).
Load necessary packages
source("../01 Data/load_packages.R",echo=TRUE)##
## > library("ggplot2")
##
## > library("gplots")
##
## > library("grid")
##
## > library("plyr")
##
## > library("RCurl")
##
## > library("reshape2")
##
## > library("dplyr")
##
## > library("tidyr")
##
## > library("jsonlite")
##
## > library("knitr")Dataframe Creations
source("../01 Data/cancer_tables.R",echo=TRUE)##
## > load("../01 Data/all_tables.Rdata")An analysis of cancer, age, death, and incidence through data wrangling.
source("../02 Data Wrangling/general_age_analysis.R",echo=TRUE)##
## > c_age <- cancer_by_age
##
## > c_age$COUNT <- as.character(c_age$COUNT)
##
## > c_age$COUNT[c_age$COUNT == "null"] <- 1
##
## > c_age$COUNT <- as.factor(c_age$COUNT)
##
## > all_age <- c_age %>% select(AGE, COUNT, EVENT_TYPE) %>%
## + filter(COUNT != "null") %>% group_by(AGE) %>% summarise(sum = sum(COUNT)) %>%
## + .... [TRUNCATED]
##
## > incidence_age <- c_age %>% select(AGE, COUNT, EVENT_TYPE) %>%
## + filter(COUNT != "null", EVENT_TYPE == "Incidence") %>% group_by(AGE) %>%
## + .... [TRUNCATED]
##
## > Mortality_age <- c_age %>% select(AGE, COUNT, EVENT_TYPE) %>%
## + filter(COUNT != "null", EVENT_TYPE == "Mortality") %>% group_by(AGE) %>%
## + .... [TRUNCATED]
##
## > all_age <- bind_rows(incidence_age, Mortality_age)
##
## > tbl_df(all_age)
## Source: local data frame [36 x 4]
##
## AGE sum cume type
## 1 <1 2555173 0.05555556 Incidence
## 2 1-4 3854005 0.11111111 Incidence
## 3 10-14 4729815 0.16666667 Incidence
## 4 15-19 6194267 0.22222222 Incidence
## 5 20-24 8932366 0.27777778 Incidence
## 6 25-29 11628032 0.33333333 Incidence
## 7 30-34 13484646 0.38888889 Incidence
## 8 35-39 17245125 0.50000000 Incidence
## 9 40-44 17571498 0.61111111 Incidence
## 10 45-49 18350855 0.83333333 Incidence
## .. ... ... ... ...
##
## > ggplot(all_age, aes(x = AGE, y = sum, fill = type)) +
## + geom_bar(stat = "identity") + scale_fill_manual(values = c("#000000",
## + "#CCCCCC") .... [TRUNCATED]
Inspecting Age and Site
source("../02 Data Wrangling/age_site.R",echo=TRUE)##
## > age <- cancer_by_age
##
## > age$COUNT <- as.character(age$COUNT)
##
## > age$COUNT[age$COUNT == "null"] <- 1
##
## > age$COUNT <- as.factor(age$COUNT)
##
## > breasts <- c("Female Breast", "Male Breast", "Male and Female Breast <i>in situ</i>",
## + "Female Breast, <i>in situ</i>", "Male and Female Breas ..." ... [TRUNCATED]
##
## > breast <- c(`Female Breast` = "Breast", `Male Breast` = "Breast",
## + `Male and Female Breast <i>in situ</i>` = "Breast", `Female Breast, <i>in s .... [TRUNCATED]
##
## > age$SITE <- as.character(age$SITE)
##
## > age$SITE[age$SITE %in% breasts] <- "Breast"
##
## > age$SITE <- as.factor(age$SITE)
##
## > MAP <- c(`70-74` = "Young Adult", `27-29` = "Young Adult",
## + `20-24` = "Young Adult", `17-19` = "Child", `7-14` = "Child",
## + `7-9` = "Chil ..." ... [TRUNCATED]
##
## > age <- age %>% select(AGE, SITE, EVENT_TYPE, COUNT,
## + RACE, SEX) %>% filter(COUNT != "null", SITE != "All Cancer Sites Combined",
## + SEX != .... [TRUNCATED]
##
## > age$AGE <- MAP[age$AGE]
##
## > tbl_df(age)
## Source: local data frame [124,488 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > total <- age %>% group_by(SITE) %>% summarise(sum = sum(COUNT)) %>%
## + arrange(desc(sum)) %>% slice(1:10)
##
## > tbl_df(total)
## Source: local data frame [10 x 2]
##
## SITE sum
## 1 Brain and Other Nervous System 16690236
## 2 Leukemias 16250318
## 3 Colon and Rectum 15496776
## 4 Non-Hodgkin Lymphoma 14243447
## 5 Lung and Bronchus 13848469
## 6 Liver and Intrahepatic Bile Duct 12923969
## 7 Stomach 12816624
## 8 Kidney and Renal Pelvis 12020378
## 9 Oral Cavity and Pharynx 11153134
## 10 Breast 11053659
##
## > total_7 <- as.character(total$SITE)
##
## > total_7
## [1] "Brain and Other Nervous System" "Leukemias"
## [3] "Colon and Rectum" "Non-Hodgkin Lymphoma"
## [5] "Lung and Bronchus" "Liver and Intrahepatic Bile Duct"
## [7] "Stomach" "Kidney and Renal Pelvis"
## [9] "Oral Cavity and Pharynx" "Breast"
##
## > child <- age %>% filter(AGE == "Child") %>% group_by(SITE) %>%
## + summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>% slice(1:7)
##
## > child_7 <- as.character(child$SITE)
##
## > yadult <- age %>% filter(AGE == "Young Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > yadult_7 <- as.character(yadult$SITE)
##
## > madult <- age %>% filter(AGE == "Middle Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1: .... [TRUNCATED]
##
## > madult_7 <- as.character(madult$SITE)
##
## > eadult <- age %>% filter(AGE == "Elder Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > eadult_7 <- as.character(eadult$SITE)
##
## > all_top <- c(child_7, yadult_7, madult_7, eadult_7,
## + total_7)
##
## > all_top <- unique(all_top)
##
## > tbl_df(age)
## Source: local data frame [124,488 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > age <- filter(age, SITE %in% all_top) %>% group_by(AGE,
## + SITE, EVENT_TYPE, RACE, SEX)
##
## > ggplot(age, aes(AGE), weight = COUNT, fill = SITE) +
## + scale_x_discrete(limits = c("Child", "Young Adult", "Middle Adult",
## + "Elder Ad ..." ... [TRUNCATED]
Some Data manipulation…
source("../02 Data Wrangling/age_site2.R",echo=TRUE)##
## > age <- cancer_by_age
##
## > breasts <- c("Female Breast", "Male Breast", "Male and Female Breast <i>in situ</i>",
## + "Female Breast, <i>in situ</i>", "Male and Female Breas ..." ... [TRUNCATED]
##
## > breast <- c(`Female Breast` = "Breast", `Male Breast` = "Breast",
## + `Male and Female Breast <i>in situ</i>` = "Breast", `Female Breast, <i>in s .... [TRUNCATED]
##
## > age$SITE <- as.character(age$SITE)
##
## > age$SITE[age$SITE %in% breasts] <- "Breast"
##
## > age$SITE <- as.factor(age$SITE)
##
## > MAP <- c(`70-74` = "Young Adult", `27-29` = "Young Adult",
## + `20-24` = "Young Adult", `17-19` = "Child", `7-14` = "Child",
## + `7-9` = "Chil ..." ... [TRUNCATED]
##
## > age <- age %>% select(AGE, SITE, EVENT_TYPE, COUNT,
## + RACE, SEX) %>% filter(COUNT != "null", SITE != "All Cancer Sites Combined",
## + SEX != .... [TRUNCATED]
##
## > age$AGE <- MAP[age$AGE]
##
## > tbl_df(age)
## Source: local data frame [54,643 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > total <- age %>% group_by(SITE) %>% summarise(sum = sum(COUNT)) %>%
## + arrange(desc(sum)) %>% slice(1:10)
##
## > tbl_df(total)
## Source: local data frame [10 x 2]
##
## SITE sum
## 1 Brain and Other Nervous System 16684764
## 2 Leukemias 16244846
## 3 Colon and Rectum 15491304
## 4 Non-Hodgkin Lymphoma 14237975
## 5 Lung and Bronchus 13842997
## 6 Liver and Intrahepatic Bile Duct 12918497
## 7 Stomach 12811152
## 8 Kidney and Renal Pelvis 12014906
## 9 Oral Cavity and Pharynx 11147662
## 10 Breast 11046819
##
## > total_7 <- as.character(total$SITE)
##
## > total_7
## [1] "Brain and Other Nervous System" "Leukemias"
## [3] "Colon and Rectum" "Non-Hodgkin Lymphoma"
## [5] "Lung and Bronchus" "Liver and Intrahepatic Bile Duct"
## [7] "Stomach" "Kidney and Renal Pelvis"
## [9] "Oral Cavity and Pharynx" "Breast"
##
## > child <- age %>% filter(AGE == "Child") %>% group_by(SITE) %>%
## + summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>% slice(1:7)
##
## > child_7 <- as.character(child$SITE)
##
## > yadult <- age %>% filter(AGE == "Young Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > yadult_7 <- as.character(yadult$SITE)
##
## > madult <- age %>% filter(AGE == "Middle Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1: .... [TRUNCATED]
##
## > madult_7 <- as.character(madult$SITE)
##
## > eadult <- age %>% filter(AGE == "Elder Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > eadult_7 <- as.character(eadult$SITE)
##
## > all_top <- c(child_7, yadult_7, madult_7, eadult_7,
## + total_7)
##
## > all_top <- unique(all_top)
##
## > tbl_df(age)
## Source: local data frame [54,643 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > age <- filter(age, SITE %in% all_top) %>% group_by(AGE,
## + SITE, EVENT_TYPE, RACE, SEX)
##
## > ggplot(age, aes(AGE), weight = COUNT, fill = SITE) +
## + scale_x_discrete(limits = c("Child", "Young Adult", "Middle Adult",
## + "Elder Ad ..." ... [TRUNCATED]
Further analysis, introducing race.
source("../02 Data Wrangling/age_site3.R",echo=TRUE)##
## > age <- cancer_by_age
##
## > breasts <- c("Female Breast", "Male Breast", "Male and Female Breast <i>in situ</i>",
## + "Female Breast, <i>in situ</i>", "Male and Female Breas ..." ... [TRUNCATED]
##
## > breast <- c(`Female Breast` = "Breast", `Male Breast` = "Breast",
## + `Male and Female Breast <i>in situ</i>` = "Breast", `Female Breast, <i>in s .... [TRUNCATED]
##
## > age$SITE <- as.character(age$SITE)
##
## > age$SITE[age$SITE %in% breasts] <- "Breast"
##
## > age$SITE <- as.factor(age$SITE)
##
## > MAP <- c(`70-74` = "Young Adult", `27-29` = "Young Adult",
## + `20-24` = "Young Adult", `17-19` = "Child", `7-14` = "Child",
## + `7-9` = "Chil ..." ... [TRUNCATED]
##
## > age <- age %>% select(AGE, SITE, EVENT_TYPE, COUNT,
## + RACE, SEX) %>% filter(COUNT != "null", SITE != "All Cancer Sites Combined",
## + SEX != .... [TRUNCATED]
##
## > age$AGE <- MAP[age$AGE]
##
## > tbl_df(age)
## Source: local data frame [54,643 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > total <- age %>% group_by(SITE) %>% summarise(sum = sum(COUNT)) %>%
## + arrange(desc(sum)) %>% slice(1:10)
##
## > tbl_df(total)
## Source: local data frame [10 x 2]
##
## SITE sum
## 1 Brain and Other Nervous System 16684764
## 2 Leukemias 16244846
## 3 Colon and Rectum 15491304
## 4 Non-Hodgkin Lymphoma 14237975
## 5 Lung and Bronchus 13842997
## 6 Liver and Intrahepatic Bile Duct 12918497
## 7 Stomach 12811152
## 8 Kidney and Renal Pelvis 12014906
## 9 Oral Cavity and Pharynx 11147662
## 10 Breast 11046819
##
## > total_7 <- as.character(total$SITE)
##
## > total_7
## [1] "Brain and Other Nervous System" "Leukemias"
## [3] "Colon and Rectum" "Non-Hodgkin Lymphoma"
## [5] "Lung and Bronchus" "Liver and Intrahepatic Bile Duct"
## [7] "Stomach" "Kidney and Renal Pelvis"
## [9] "Oral Cavity and Pharynx" "Breast"
##
## > child <- age %>% filter(AGE == "Child") %>% group_by(SITE) %>%
## + summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>% slice(1:7)
##
## > child_7 <- as.character(child$SITE)
##
## > yadult <- age %>% filter(AGE == "Young Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > yadult_7 <- as.character(yadult$SITE)
##
## > madult <- age %>% filter(AGE == "Middle Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1: .... [TRUNCATED]
##
## > madult_7 <- as.character(madult$SITE)
##
## > eadult <- age %>% filter(AGE == "Elder Adult") %>%
## + group_by(SITE) %>% summarise(sum = sum(COUNT)) %>% arrange(desc(sum)) %>%
## + slice(1:7 .... [TRUNCATED]
##
## > eadult_7 <- as.character(eadult$SITE)
##
## > all_top <- c(child_7, yadult_7, madult_7, eadult_7,
## + total_7)
##
## > all_top <- unique(all_top)
##
## > tbl_df(age)
## Source: local data frame [54,643 x 6]
##
## AGE SITE EVENT_TYPE COUNT
## 1 Elder Adult Colon and Rectum Incidence 9993
## 2 Elder Adult Colon and Rectum Incidence 9992
## 3 Elder Adult Corpus and Uterus, NOS Mortality 999
## 4 Elder Adult Oral Cavity and Pharynx Incidence 999
## 5 Elder Adult Corpus and Uterus, NOS Mortality 999
## 6 Middle Adult Liver and Intrahepatic Bile Duct Mortality 999
## 7 Young Adult Lung and Bronchus Incidence 999
## 8 Middle Adult Colon and Rectum Mortality 999
## 9 Middle Adult Prostate Incidence 9982
## 10 Elder Adult Pancreas Incidence 998
## .. ... ... ... ...
## Variables not shown: RACE (fctr), SEX (fctr)
##
## > age <- filter(age, SITE %in% all_top) %>% group_by(AGE,
## + SITE, EVENT_TYPE, RACE, SEX)
##
## > ggplot(age, aes(AGE), weight = COUNT, fill = SITE) +
## + scale_x_discrete(limits = c("Child", "Young Adult", "Middle Adult",
## + "Elder Ad ..." ... [TRUNCATED]
Let’s now focus on geography. Alaska and Hawaii have been left out to improve the aesthetics of the following visualizations.
Comparing types of cancer occurence in each state
Non-Hodgkin Lymphomas are by far the most common in each state. Interestingly, each state has very similar proportions of cancer types, leading one to believe that the form of cancer you are diagnosed with is not related to region.
Comparing cancer occurence by race in each state.
Notice how white people form the majority in every state, nearly equal in size to the section for all races.
Comparing cancer occurence by racial minorities in each state
Taking out white people and all races leaves us with minority categories. These statistics appear to follow the demographic ratios that are prevalent in these regions, with a more black concentrated South East and a more Hispanic concentration to the West. Apparently, minorities either do not exist or do not get cancer in the Central North.
Let’s take another look at the most common cancers.
Common cancers
Here is a circle chart comparing the proportion of cancer cases with each type of cancer. While the color is based on absolute count, the size is based on the percent occurence using the Count/Population calculated field below:
Using this calculated field allows us to factor in the populations of each region in question and to summarize cancer occurence by these regional percentages.
By percentage, lung, prostate, and breast cancers are very similar and in lead as the most common cancers. By count, however, lung cancer is far more prevalent.
Common cancers by sex
While female breast cancer and prostate cancer are leading forms of cancers in their respective sex groups, both sexes suffer second most from lung cancer.
Common cancers by age
As is common knowledge, cancer becomes a growing concern as one ages. It appears that Female Breast Cancer is one form of cancer that women must worry about for the majority of their lives.
Common cancers by year
Though the proportion of cancers to each other are staying constant, the number of cancer incidents are increasing, as indicated by the increasing darkness of lung cancer and breast cancer in particular.
We will now focus on age range
Types of cancers for ages by race and sex
White people have the highest occurence of cancer by far, most likely because they are the most populous race in the United States, but several other factors are likely involved in these high numbers. If also appears that males start to be diagnosed with cancer more than women around ages 55-84
Types of cancers for children by race and mortality
Here are the cancers that children are diagnosed with and whether they are fatal or not. Thankfully, it appears that most children with cancer will recover. Early childhood is plagued by leukemias and brain cancers. It is not until age 15 that we begin to see a large variety of cancers such as thyroid, testicular, and melanomas.
Types of cancer for females by race and mortality
Age 35 is when females start to become increasingly diagnosed with cancer. Highlighted here are some of the more commonly associated female cancers. While breast cancer occurs the most, lung cancer is the most lethal.
Types of cancer for males by race and mortality
Males start to become increasingly diagnosed with cancer around age 55, where the most common to be diagnosed is prostate cancer. Like women, lung cancer is the most lethal for men, though prostate cancer becomes increasingly more lethal as the person gets older.
Lets look at some crosstabs in Tableau
First, we’ll look at the Rank crosstab. The function ranks the largest number of counts per state per year and partitions by sex, race, and cancer site. Only 5 states are included to ease viewing: California, Texas, Illinois, Florida, New York. California consistently produces the most cancer-caused deaths year after year.
And the field calculation:
Next, we’ll look at the Nth crosstab. This function returns the nth largest count value, where n can be any positive integer. Here the counts are partitioned by age, sex, behavior, and cancer site. The Nth parameter allows us to select which count we wish to see.
And the field calculation:
We’ll also look at the Maximum & Difference crosstab. The max and diff functions calculate the maximum number of counts and the difference between the maximum and the current value respectively. These crosstab columns are partitioned by Age group and year. We also have counts for each Iccc group. As can be expected, we see Leukemia as generally having the smallest difference from the maximum.
And the field calculations:
Finally, let’s look at a cumulative distribution crosstab. This function gives the probability that a value picked at random will be below another number. In this case, we look at cancer in children at various sites. We partition count by age group, year, and cancer site. The distribution is represented in the final column. The built-in percentile quick calculation was used here. A quick observation that fits this data: A child is much more likely to develop Leukemia than kidney-related cancer.
